Work files are files to which data can be written and from which
data can be read by Natural programs. They are used for intermediate storage of
data and for data exchange between programs. Data can be transferred from or to
a work file by using the Natural statements READ WORK FILE and
WRITE WORK
FILE.
This document covers the following topics:
Using the Configuration
Utility or the DEFINE WORK
FILE statement, you can assign names (including the path) for
up to 32 work files.
The maximum number of work files that can be used depends on the
setting of the parameter WORK.
If you run a program which uses a work file for which a name and
path has not been assigned, Natural automatically creates the file name and
writes the work file into the
temporary
directory specified in the local configuration file. The name of
such a file consists of the specified work file number and an arbitrary number
assigned by the operating system. The generation of the work file name is based
on an algorithm which tries to generate a unique name. Depending on the Natural
parameter TMPSORTUNIQ,
the naming convention may vary. If work file names are referenced from outside
Natural, it is recommended that you specify the names explicitly to avoid
problems identifying the files.
The following topics are covered below:
In the Configuration Utility, the work file names are assigned
in the category
Work
Files of a parameter file. The above mentioned
parameters WORK and TMPSORTUNIQ
can also be found in this category. See
Work
File Assignments in the Configuration
Utility documentation for further information.
Tip:
Locate the work file assignments by searching for
"Work Files". See
Finding a
Parameter in the Configuration
Utility documentation for further information.
The following topics are covered below:
Work files can also be defined by using Windows environment
variables. Once you have defined
the work file names in the parameter file, the work file names can be set
without further change to the parameter file. For example, when you specify the
following name for a work file in the parameter file (or in a DEFINE WORK
FILE statement):
%Natural%\%myfile%
and assume the following settings in your operating system:
set Natural=D:\natural set myfile=sub\test
this will expand into the following file name:
D:\natural\sub\test
Names of environment variables are delimited by special characters. A left-hand delimiter is to the left of a variable, a right-hand delimiter is to the right.
For example, the string %TEMP% identifies an environment variable named
TEMP; % is used as both the left-hand and right-hand delimiter.
Valid delimiters are:
| Type of Delimiter | Valid Delimiters |
|---|---|
| Left-hand delimiter |
% |
| Right-hand delimiter |
% |
Note:
The end-of-string mark is by default a right-hand
delimiter, i.e. %TEMP is recognized as an environment variable
named TEMP.
Although "%" is the only valid
left-hand delimiter for environment variables in Windows, Natural for Windows
allows "%" and "$" as
left-hand delimiters in order to preserve upward compatibility with previous
versions. This setting allows UNIX-like work file name assignments in a Windows
session. $TEMP is recognized in Natural for UNIX as well as in
Natural for Windows as the environment variable
TEMP.
Example:
The following lines of Natural code are interpreted as being the same:
DEFINE WORK FILE 1 '$TEMP\myfile.dat'
and
DEFINE WORK FILE 1 '%TEMP%\myfile.dat'
TEMP is recognized as an environment variable.
The string $TEMP (or %TEMP%) is replaced at runtime
by the contents of the environment variable TEMP.
A dollar sign ($) in a file name has two meanings:
If the dollar sign appears on the left or in the middle of a string embedded in delimiters, it will be interpreted as the left-hand delimiter of the environment variable being used. All characters following the left-hand delimiter up to the right-hand delimiter or EOS are considered to be the name of an environment variable.
If the dollar sign is the last character of a string, it is not considered to be a delimiter character. It is a part of the string scanned.
Example:
The following line of Natural code does not result in an error:
DEFINE WORK FILE 1 '\\MYPC\C$\myfile.dat'
\\MYPC\C$ is considered to be a default share.
C$ is a valid directory.
However, the following line of Natural code may result in an error, depending on whether A has been defined or not:
DEFINE WORK FILE 1 '\\MYPC\C$A\myfile.dat'
A is interpreted as an environment variable since
it is preceded by a dollar sign. If A has not been defined, an
error will occur. If A has been defined, an error does not
occur.
You can also define work files with the application programming
interface USR1050N in library SYSEXT.
The format of a work file depends on the work file type that has been defined. Different work file formats are available. Natural recognizes the format by checking the file name and its extension:
file-name.extension
where file-name can have a maximum of 8 characters and extension can have a maximum of 3 characters.
The work file formats are:
See also Work Files and Print Files in the Unicode and Code Page Support documentation.
Possible type: SAG.
This format, which is specific to Software AG, is the preferred format since it can be used with all data types. However, it is not portable across platforms with different endian modes.
Each record that is written is preceded by two bytes which contain the length of the record. The length itself is written in a platform-specific form.
To define binary format for a work file, use a file name with a period and the extension SAG (for example, <file-name>.SAG).
Possible types: ASCII and ASCII compressed.
Since each written record is terminated with a carriage return and line feed (CR/LF), ASCII format is only recommended for alphanumeric data.
To define ASCII format for a work file, enter either a file name with a period and any extension except SAG and NCD (for example, <file-name>.<ext>), or a file name with a period and without an extension (for example, <file-name>).
Possible type: Entire Connection.
The product Entire Connection uses two files: a data file which contains the actual data and a format file which contains formatting information about the data in the data file.
Natural automatically generates the corresponding format file for the type Entire Connection. The format file has the same name as the data file, however the extension is NCF. For detailed information on the content of a format file with the extension NCF, see the Entire Connection documentation.
To define Entire Connection format for a work file, enter a file name with a period and the extension NCD (for example, <file-name>.NCD).
You can read/write work files in Entire Connection format directly from/to your local disk.
See also Special Considerations for Work Files with Extension NCD.
Notes:
RECORD option of the
READ WORK FILE
statement is not available for reading work files of format Entire
Connection.
Possible type: Portable.
The type Portable performs an automatic endian conversion of a work file when the work file is transferred to a different machine. For example, a work file written on a PC (little endian) can be read correctly on an RS6000 or HP machine (big endian). The endian conversion applies only to field formats I2, I4, F4, F8 and U. The floating point format is assumed to be IEEE. There are, however, slight differences in IEEE floating point representation by different hardware systems. As a rule, these differences apply only to infinity and NaN representations, which are normally not written into work files. Check the hardware descriptions if you are uncertain.
The files are always written in the machine-specific representation, so that a conversion is performed only if the file is read by a machine with different representation. This keeps performance as fast as possible.
There are no other conversions for this format apart from the conversions mentioned above.
When a READ WORK
FILE statement is used for a dynamic variable, the variable
is resized to the length of the current record.
Possible type: Unformatted.
The type Unformatted reads or writes a complete file with just one dynamic variable and just one record (for example, to store a video which was read from a database). No formatting information is inserted; everything is written and read just as it is.
Possible type: CSV (comma-separated values).
Note:
If you want to use the work file type CSV, you have to
recatalog your sources using the CATALOG or
STOW
command. It is not possible to use the work file type CSV with generated
programs of Natural Version 4.
The Natural fields are stored in a CSV work file as described below.
In the first step, the internal field data is converted into a readable format:
The field data of the internal Natural data formats B (binary), O (object handle), G (GUI handle) and C (attribute control) is copied to the record without field conversion. The data is taken as it is.
The field data of the internal Natural data format A
(alphanumeric) is converted into the specified work file code page (see
Work
Files in the Configuration Utility
documentation). If no work file code page is specified in the Configuration
Utility, the default code page which is defined with the parameter
CP is used
and no conversion is done.
The field data of the internal Natural data format U
(Unicode), is converted into the specified work file code page (see
Work
Files in the Configuration Utility
documentation) or, if no work file code page is specified, into the default
code page which is defined with the parameter CP.
The values of the internal Natural formats D (date) and
T (time) are converted into an alphanumeric output format. The
DTFORM
parameter is evaluated so that the user-specified date and time format is
used.
The internal field values of the numeric types are converted into an alphanumeric output format.
In the second step, the field data in readable format is copied to the CSV work file record. The fields in the work file are separated by the specified separator character. If a field contains special characters, the field is delimited by double quotes. Each written record is terminated with a carriage return and line feed (CR/LF).
If you have defined that a header with the Natural field names is to be written to the work file (see Work File Assignments in the Configuration Utility documentation), the following applies:
With the WRITE WORK
FILE statement, a header line containing the field names of
the first written record is stored in the first line of the work file. If
subsequent CSV records contain a different number of fields, it may be possible
that the header line does not correspond to these subsequent CSV records.
With the READ WORK
FILE statement, it is assumed that the first line of the CSV
work file is the header line. Therefore, the first line is skipped (that is:
the record data in the first line is not returned).
If files with the extension NCD are created
by Entire Connection and are then read into Natural via the READ WORK
FILE statement, it is required that the Entire Connection option
Keep trailing blanks is activated in the session
properties. See your Entire Connection documentation for further
information.
Note:
When you create an NCD file using Entire Connection and load
this file using the Object
Handler, you may receive an error indicating that the source control
record is missing. To avoid this, make sure that the option Keep
trailing blanks is active when you create the NCD file.
The following considerations apply for work files in Entire Connection format:
If an NCD file is read with a READ WORK FILE
statement and the corresponding NCF format file is not available or contains
invalid information, the NCD file is assumed to be an ASCII work file.
When the APPEND attribute is used to append data
to an NCD file, the record layouts (that is: the field format and length
information which is written to the NCF format file) of the old and new data
must match. If the record layouts are different, an error occurs during
runtime.
The maximum work-file record size for WRITE WORK FILE
VARIABLE that can be handled by Entire Connection is 32767 bytes.
If you have "old" work files with the extension NCD, the extensions must be changed.
Each of the following profile parameters must be set to the same value for both read and write operations:
DC (decimal character)
IA (input assign
character)
ID (input delimiter
character)
Remember that the range of possible values for floating point variables on a mainframe computer is different from that on other platforms. The possible value range for F4 and F8 variables on a mainframe is:
±5.4 * 10-79 to ±7.2 * 1075
The possible value range on most other platforms for F4 variables is:
±1.17 * 10-38 to ±3.40 * 1038
The possible value range on most other platforms for F8 variables is:
±2.22 * 10-308 to ±1.79 * 10308